Bing lets you turn off AI answers by adding “-ai” to a search

DuckDuckGo has a No-AI address, Kagi offers controls, and Google and Bing have workarounds, but the actual…

Bing lets you turn off AI answers by adding “-ai” to a search

The short version

  • DuckDuckGo offers a dedicated No-AI entry point, while Kagi provides settings to disable access to AI search features.
  • Google’s Web search URL requests a filtered view, while Bing documents adding -ai to turn off AI answers.
  • Avoiding generated answers does not establish privacy or result quality; browser defaults, mobile settings and service policies still need checking.

I asked for links and my search engine brought a monologue

My search engine’s job is to find the restaurant, then let me read the damn menu. Dinner shouldn’t require browser configuration.

Bar chart comparing current figures against their baselines: Hydrogen task success 57 % versus 18 %, Share of citation IDs that were fabricated 0 % versus 6 %, Real-research requests that received at… 84 % versus 100 %, ChatGPT sessions accompanied by a… 5 % versus 31 %.

When I search for “search engines that don't use AI,” I want the source without a generated answer taking over. I use AI elsewhere. I also own a blender; I don’t want every meal served through it.

DuckDuckGo has a dedicated No-AI address. Kagi lets me disable AI features in search. Google and Bing offer other routes. Before declaring victory, I need to check what my address bar actually uses.

I’d choose the entry point before changing search engines

I’d start with DuckDuckGo No-AI. According to TechCrunch’s November coverage, it removes the service’s other AI features and filters AI-generated images by default. I still expect synthetic images to slip through; a filter guarantees nothing about an individual picture.

The address matters: choosing ordinary DuckDuckGo in a browser menu doesn’t necessarily use its No-AI site.

I distinguish these options by where the preference lives. DuckDuckGo offers a dedicated entry point. Kagi’s AI settings let me disable access to AI features in search. Carpathian describes ranked pages from its own index. It says any answer above them quotes the best-matching page and names its source. That gives me a shorter path to inspect the original, not protection against missing context, wrong sources or machine-written pages.

None of this research includes a technical audit establishing that any named engine avoids machine learning throughout its search pipeline. That promise needs more than a clean screenshot.

Kagi’s ad-free service makes my paid-search shortlist. Its changelog credits community feedback with helping motivate AI controls, while the company keeps investing in AI elsewhere. I like that: customers asked for a boundary, and got one.

Kagi’s November 5 announcement offered 50 guest searches without an account, then 100 more after registration. Those were introductory allowances, not a permanent free plan. I’d check the current offer.

DuckDuckGo’s dedicated option is free, according to The AI Future’s comparison. Price says nothing about data handling; I still need the privacy terms.

A cobalt smartphone with a cream search symbol lies exposed in terracotta excavation layers, beside a small blank menu.

I can keep Google or Bing, with some browser plumbing

I can select Web, sometimes under More, after a Google search. Remembering every time is where my preferences collide with my personality.

NBC10 Boston’s November guide describes a desktop Chrome workaround: make a custom Google Web search URL the default. That automates the filtered route, not an account-wide Google off switch.

Chrome sends address-bar searches through a saved URL template, replacing %s with my query. Adding udm=14 to Google’s regular search URL requests Web view. Making that template my default automatically requests the filtered page. Google’s regular homepage takes a separate route and can show a different presentation. My phone needs its own setup; desktop Chrome proves nothing about the browser I open outside a restaurant.

NBC10 Boston’s desktop Chrome procedure:

  • Open Settings → Search engine → Manage search engines and site search.
  • Under Site search, select Add and enter a name and shortcut.
  • Replace the regular Google search URL with https://www.google.com/search?q=%s&udm=14 to request Web results.
  • Save the entry, open its menu, and choose Make default.

I finish with an address-bar search. Browser settings happily accept changes I later discover I never used.

Bing offers a query-level option. Microsoft’s November 11 post announcing AI-answer controls documents appending -ai to switch them off versus an ordinary query. Its example: why is the sky blue -ai.

Professor H’s November 10 test found no Copilot Search boxes across twenty queries with -ai, versus twelve across the same twenty without it. Conditions matter: desktop browser, signed out, US English.

That tells me nothing about signed-in accounts, phones or apps. Professor H also flags a possible term-exclusion effect: Bing uses minus signs to exclude search terms. Researching AI itself needs an extra check.

Microsoft’s alternative, its preview AI Search Choice extension for Chrome and Edge, also makes Bing the default. Quite a commitment if I only wanted an answer box gone.

In its November 11 on/off-controls announcement, Microsoft Bing explained:

We're proud of the AI experiences we're building in Bing. Being able to turn AI answers off when you want to isn't a step back from AI. It's a step toward putting you in charge of how you use it.

I welcome that control. But I understand a school administrator’s reservations: neither a signed-out desktop test nor the extension documentation establishes compliance with a particular school’s managed-device requirements.

Google says search and AI are deeply integrated; its stated school controls don’t let schools keep Google Search while disabling built-in AI features. DuckDuckGo argues there’s no technical reason Google couldn’t offer that choice. Its No-AI service demonstrates another product decision, not how Google’s systems work. The sources leave a comprehensive administrator switch’s feasibility and operating cost unresolved.

My quiet results page still has to survive the AI economy

I’d love to judge an engine entirely by its results page. Brave’s account of scraping traffic ruins that shortcut.

Kanishka Narayan described the economic stakes of developing technology:

For too long, Britain has produced some of the world’s best scientific ideas, only to see too much of the technology, industry and wealth they create built elsewhere.

Brave attributes increased scraping against its service to Google’s GOTO anti-scraping change. I have no independent confirmation, but Brave describes consequences even for people who never use generated answers.

According to Brave, extra scraping prompted more aggressive bot protection, slowing searches and increasing CAPTCHAs for real people. Brave then developed encrypted result-link cloaking to deter scraping. It says this protects links without revealing users’ clicked destinations to the company. Without JavaScript, however, a conventional-redirect fallback can reveal the clicked URL to Brave: the disclosure encryption was designed to avoid. My AI-off preference touches none of these infrastructure decisions.

Kat Jones described AI datacentres’ power requirements:

Hyperscale AI datacentres simply cannot run on renewable power alone, they absolutely need a highly reliable source of grid power.

I take the operator’s problem seriously: leaving scraping alone also costs users search speed. Brave’s stated goal is to deter scrapers while preserving click privacy. Its JavaScript fallback is exactly the qualification I want before recommending a service.

I apply the same product-by-product scrutiny to privacy promises. Mojeek’s November privacy policy says standard website logs include search terms and remain for 90 days; API customers can choose zero retention for API queries.

That changes what I can tell a friend. An API customer’s promise doesn’t automatically follow me into the website’s search box.

Before switching, I need click-handling details and the policy for my actual service. For Brave, that means checking the JavaScript fallback. For Mojeek, website terms, not a reassuring API claim.

I’d make the new default earn its place

A missing restaurant website irritates me more than an unfashionable search logo. I’m Italian; menu access is a serious performance requirement.

My research includes no controlled comparison of these answer-free options across representative everyday searches. I can explain controls, not crown a relevance winner, however much I enjoy calmer results.

I first check what each service can retrieve. An index holds searchable page information; ranking can’t conjure up a page that never entered it. Refreshes matter when business hours change or stories develop. Services using outside providers can share underlying material but rank it differently. Independent indexes make their own collection decisions, with different potential gaps. That’s why I’d test familiar queries before moving my daily searches.

According to Damiano Spina’s November analysis in The Conversation, DuckDuckGo combines sources including Bing results and its own crawler’s material. The Knowledge Engineers’ Kagi profile likewise describes Kagi’s own index mixed with third-party search providers.

A new logo doesn’t prove I’ve acquired a separate view of the web.

Carpathian says its index is smaller than Google’s, with no image, video or shopping search. Those omissions matter if I use those modes, but say less about its web results’ usefulness for my queries.

I’d try ordinary searches: a local business whose website I know, a recent development, then an obscure query to expose gaps. For shopping, “best budget smartphone” would test whether I reach useful reviews, not just admire empty space above results.

I haven’t run that comparative benchmark. My recommendation remains a setup to try, not a verdict on its results.

I’d make it my actual default for a week, with a larger engine bookmarked for misses. That tests normal behavior. Opening a folder of alternatives whenever I remember I’m evaluating search engines mostly tests my memory.

I’d check my phone separately. A beautiful desktop setup earns little credit if my next distracted search bypasses it.

For schools, I can’t justify another tempting leap: the sources don’t establish whether hiding generated answers improves learning outcomes versus guided AI assistance. My preference for original sources can’t settle that.

My requirement is smaller. Hungry outside a restaurant, I want my usual search gesture to reach its website without more settings archaeology. If the menu turns out to be a PDF from 2019, I’ll finally be annoyed at the restaurant again.

Frequently asked questions

Which search engines let you search without AI answers?

DuckDuckGo offers a dedicated No-AI address, and Kagi provides settings to disable access to AI features in search. Google offers a Web view, while Bing documents a query-level control for AI answers. These options do not establish that the engines avoid machine learning throughout their search pipelines.

How do you make Google Web search the default in desktop Chrome?

Desktop Chrome can use a custom search entry with the URL https://www.google.com/search?q=%s&udm=14. The %s placeholder receives the query, and udm=14 requests Web view. Saving that entry and making it the default routes address-bar searches through Web view. This does not create an account-wide switch or configure a phone’s browser.

Does turning off AI search answers make a search engine more private?

Turning off generated answers does not establish a search service’s privacy practices. Brave describes a JavaScript-free fallback that can reveal clicked URLs to the company. Mojeek’s cited website policy retains standard logs, including search terms, for 90 days, while its API customers can choose zero retention for API queries.

Sources

Related reading

Luca

Luca

Luca by the way is the personal blog of Los Angeles based entrepreneur Luca Capula. A true Italian who lives between Torino and LA.

More posts →